Statistics 351 Fall 2008 Midterm #1 Solutions

1. (a) By definition,




 = ey ,
dx = e


fY (y) =
y

y > 0.

Note that Y Exp(1) so that E(Y ) = 1.

1. (b) By definition,

ex dy = xex ,

fX (x) =

x > 0.

Note that X (2, 1).

1. (c) By definition,
fY |X=x (y) =

fX,Y (x, y)
ex
1
= x = ,
fX (x)
xe
x

0 < y < x.

Note that Y |X = x U (0, x).

1. (d) By definition,
Z

E(Y |X = x) =

Thus, we conclude E(Y |X) =

y fY |X=x (y) dy =
0

x
1
dy = .
x
2

X
.
2

1. (e) Using (d) we find


E(Y ) = E(E(Y |X)) = E

X
2

1
= E(X).
2

However, from (a) we know that E(Y ) = 1 and so

E(X) = 2.
1. (f ) If a < 1, then we find
ZZ
P {Y < aX} =

ax

fX,Y (x, y) dx dy =

e
0

{y<ax}

Z
dy dx =

axex dx

= a(2)
= a.
1. (g) If U = X + Y and V = 2Y , then solving for X and Y gives
X=U

V
2
1

and Y =

V
.
2

The Jacobian of this transformation is



 x x 

 

 u v  1 1/2
=

J = 
 0 1/2  = 1/2.
y
y




u v
Therefore, we conclude
fU,V (u, v) = fX,Y (u v/2, v/2) |J| = eu+v/2

1
1
= eu+v/2
2
2

provided that 0 < v < u < . The marginal for U is given by

u
Z
Z



1 u u v/2
1 u
2ev/2  = eu (eu/2 1) = eu/2 eu
fU (u) =
fU,V (u, v) dv = e
e dv = e
2
2

0
0
provided u > 0.
2. (a) Notice that

Xn+1 =

1p
p

Sn+1


=

1p
p

Sn +Yn+1


=

1p
p

 Sn 

1p
p

Yn+1

Therefore,
!
S 
Y
1 p n 1 p n+1
E(Xn+1 |X0 , . . . , Xn ) = E
|X0 , . . . , Xn
p
p
!

S
Y

1p n
1 p n+1
=
E
|X0 , . . . , Xn
p
p

S
Y !

1p n
1 p n+1
=
E
p
p


where the second equality follows from taking out what is known and the third
equality follows from the fact that Y1 , Y2 , . . . are independent. We now compute


Y !
1

1
1p
1 p n+1
1p
=p
+ (1 p)
= (1 p) + p = 1
E
p
p
p
and so we conclude

E(Xn+1 |X0 , . . . , Xn ) =

1p
p

Hence, {Xn , n = 0, 1, 2, . . .} is, in fact, a martingale

Sn
= Xn .

2. (b) Since Xn is a martingale, we know that E(Xn+1 |Xn ) = Xn . Therefore, using properties
of conditional expectation we find
E(Xn+1 ) = E(E(Xn+1 |Xn )) = E(Xn ) = = E(X0 ).
Since S0 = 0 we see that X0 = 1 so that E(X0 ) = 1 and therefore E(Xn ) = 1 for all
n = 0, 1, 2, . . ..
3.

If U = Y X and V = X, then solving for X and Y gives

X=V

and Y = U + V.

The Jacobian of this transformation is



 x x 

 

 u v  0 1
=

J = 
 1 1 = 1.
y
y




u v
Therefore, we conclude
fU,V (u, v) = fX,Y (v, u + v) |J| = c(c + 1)(b a)c uc2 1 = c(c + 1)(b a)c uc2
provided that < v < a and < b v < u < .
Z

You should check that

4.

c(c + 1)(b a)c uc2 du dv = 1.

bv

By the law of total probability,

Z
Z
fX (x) =
fX|Y =y (x)fY (y) dy =

y q q1 yx ap p1 ay
x e

y e
dy
(q)
(p)
0
Z
ap
q1
=
x
y p+q1 ey(x+a) dy.
(p)(q)
0

We now recognize
Z

y p+q1 ey(x+a) dy

as a gamma function. That is, make the change of variables u = y(x + a) so that
du = (x + a) dy and
p+q1

p+q Z
Z
Z 
u
1
p+q1 y(x+a)
u du
y
e
dy =
e
=
up+q1 eu du
x+a
x+a
x+a
0
0
0
(p + q)
=
.
(x + a)p+q
Hence, we conclude that
fX (x) =

ap
(p + q)
(p + q) ap xq1
xq1
=
, 0 < x < .
(p)(q)
(x + a)p+q
(p)(q) (x + a)p+q

This is sometimes called a type II generalized Pareto density with parameters q, p, a.